Unencrypted login | home

Program Information

Diffusion Weighted MRI of Esophageal Cancer


J Lagendijk

A van Lier, G Meijer, M Moerland, F Lever, O Reerink, M Vulpen, J Lagendijk*, I Lips, C van den Berg, UMC Utrecht, Utrecht, The Netherlands

TH-C-WAB-9 Thursday 10:30AM - 12:30PM Room: Wabash Ballroom

Purpose: To investigate the potential of diffusion weighted MRI for esophageal cancer imaging for treatment volume definition, and to determine the need for geometrical correction of DWI images.

Methods: Three patients with esophageal cancer were included in this study. A heavily diffusion weighted free-breathing DWI scan (b=800 s/mm2) with echo-planar readout was optimized to reduce the effect of field distortions. Additionally, a B0 map (main magnetic field map), a T2w MRI and a CT were obtained and co-registered. Two radiation-oncologists delineated the primary tumour on the T2w image. High contrast between restricted (= tumor) and less restricted diffusion enabled definition of the tumour volume by thresholding. Geometrical inaccuracy of DWI images arises from inhomogeneities in the main magnetic field and the low bandwidth in the phase encoding direction, therefore inaccuracies can retrospectively be corrected using the B0 map. The dice coefficient between the delineated tumor volume and the original or correct DWI-based tumor volume was determined.

Results: In all patients, a region with restricted diffusion was observed in the esophagus. In one of the patients, also restricted diffusion was observed in adjacent (mediastinal) lymph node(s). Geometrical correction of the DWI images increased dice coefficient between the delineated volume and the thresholded volume (before: 0.75, 0.71, 0.74, after: 0.86, 0.79, 0.9,. respectively).

Conclusion: DWI of the esophagus shows excellent contrast between healthy and suspicious tissue. This enabled definition of the tumour volume using a threshold. After geometrical correction the tumour volume was in closer agreement with the delineated volume. Geometrical correction may be also useful for direct comparison of (co-registered) DWI, PET and CT images. In future, accuracy and robustness of geometrical correction will be studied in a larger patient group. Furthermore, the use of DWI for treatment reponse monitoring and detection of metastatic mediastinal lymph nodes will be investigated.


Contact Email: